Grit builds AI systems for companies and individuals. Our product, RADAR, runs AI coding agents and closes a task only after it re-runs the acceptance test itself.
Observed, not a controlled comparison
A claim alone does not close a task. See the full record
Grit internal ledger, snapshot 28 Sep 2026 06:00 UTC
How we work
Every task gets its acceptance test before any work begins.
Example. Nothing runs here.
TaskFix the redirect after sign-in
Start stays locked until the task has a test anyone can re-run.
The agent that did the work does not close it. The test is run again.
Example. Nothing runs here.
TaskFix the redirect after sign-inOpen
“Done. All tests pass.”
Locked: the agent that did the work cannot close it.
$ pytest tests/test_login.py -k redirect Open: waiting for the checker.
Every figure we publish comes from a record that names its source and date.
A real figure, not an example
From Grit's own work · Observed, not a controlled comparison
In Grit's own work (20-28 Sep 2026), workers running one coding agent reported a task as finished 1,608 times (1,457 tasks); Grit's immediate re-run of the acceptance commands on the same server failed on 74 of those reports (4.6%). A claim alone does not close a task.
Grit internal ledger, snapshot 28 Sep 2026 06:00 UTC
The work is designed to run on more than one family of AI models.
Example. Nothing runs here.
TaskFix the redirect after sign-in
AgentClaude Code
Testpytest tests/test_login.py -k redirectSame test
The agent, and the model family behind it, can change. The test that closes the task does not.
Our product
When an agent reports that a task is done, RADAR runs the task's acceptance test again. A passing test closes the task. Every step is written to a ledger you can read.
Example
Example tasks and output, written for this demo. Nothing runs here.
Add a discount code to checkout
Acceptance test
pytest tests/test_checkout.py -k discount Failure attached an expired code was still accepted
Open: the agent is working on it.
Ledger
Nothing recorded yet.
What we do
Designed and built with you, from the first draft to the live site: accessible, quick to load, and in the languages your visitors read.
For companies and individuals
Repetitive work handled by AI, the same steps done the same way each time. Every result is checked before it is used.
For teams whose work repeats
Agents that carry a piece of work from start to finish, built the way RADAR works: a task closes only when its test passes again.
For work with a clear definition of done
Scattered data brought together in one place and cleaned, so it is reliable enough to work with and to decide on.
For organisations whose data lives in many places
The company